Photoshop has learned how to automatically restore old and damaged photos

Gif: Adobe Photoshop / YouTube

There’s a new reason to start digitizing that aging collection of slowly deteriorating old snapshots in the family album. Digitally restoring old photos is often a process that can take hours, even for a skilled photo editor, but a new AI-powered filter coming to Photoshop looks handle most of the work with a single click.

Of all the neural network applications of recent years, some good ones, some badOne that has resonated strongly with the average consumer is the use of these tools as a way to connect with our past and family members who are no longer with us. Hollywood is already betting on this idea, as evidenced by Harold Ramis’ ghostly cameo in Ghostbusters: Afterlifebut that was mostly accomplished with complex (and expensive) visual effects work.

Last year, MyHeritage revealed the Deep Nostalgia tool which automatically animates old photos and turns them into short videos, making them look like home movies from distant relatives, even though they died long before smartphones and camcorders existed. But long before Deep Nostalgia arrived, the most talented photo editors were restoring damaged and faded old photos, and even colorizing images that only existed in black and white, using tools like Photoshop. The results are often amazing and can make pictures of your relatives look even more realistic, but the process requires a lot of skill and practice if you want convincing results.

New Photo Restoration Neural Filter in Photoshop

Adobe has been adding automated tools to Photoshop for a while now. Powered by the Sensei AI platform, they make mundane tasks like sky replacements and even the selection of complex objects are much easier, but soon the automated toolset will expand to photo restorations (not yet available in Photoshop 2022). Adobe teased the new feature in a short video shared on YouTube, and in addition to removing dirt, scratches, creases, and tears, the tool appears to smooth people’s skin, remove unwanted textures and noise, and add a hint of color. even on blank images and black or sepia tones.

Adobe also demonstrated that the new neural filter works with existing tools, such as contextual fill to restore missing sections of a photo because they were ripped out. The results aren’t as perfect as some of the manual photo restorations we’ve seen done by Photoshop masters (the filter is still in beta), but the results are still surprisingly good. No word on how much control Adobe plans to give users for neural filter processing (we probably wouldn’t be as aggressive with skin smoothing) or if it will remain a one-click feature when it becomes publicly available.